---
title: "Teammates can be shared now"
excerpt: "The researcher one person hires becomes something the whole company can staff work to. Headcount that compounds."
author: naomi
editedBy: govind
category: Thinking
date: "2026-07-27"
image: "/cover-samples/b-weave.png"
---

When a great employee joins a company, everyone benefits, but nobody would say the employee is *shared*. Their attention divides. An hour spent helping sales is an hour not spent helping marketing. Every organization ever built has been shaped by that constraint, so deeply that we've stopped seeing it as a constraint at all. Departments, headcount planning, the political fight over whose project gets the good engineer, all of it is rationing attention because attention was the thing that couldn't be copied.

AI teammates break that constraint, and almost nobody has updated their org thinking to match.

## What sharing actually means

Not handing someone your login. Sharing a teammate means the agent one person hired, with its role, its memory, and its accumulated context, becomes addressable by others on the team. Your co-founder messages it directly. The marketing lead hands it work. It's the same teammate, same knowledge, same standards, now holding objectives from more than one person, without the queue that forms outside a human expert's door.

The practical texture is mundane in the best way. The research teammate someone set up for sales runs a competitive brief for marketing on the same afternoon. The Chief of Staff one founder onboarded becomes the whole company's front door for scheduling and follow-ups. Nobody asks anyone to forward a chat tab, because there's no tab. There's a colleague, [in the same workspace as everyone else](/blog/your-agents-belong-where-your-team-works).

> Human expertise queues. Shared teammates don't. That single difference quietly deletes one of the oldest constraints in organizational design.

## Why this needed a roof first

Sharing is impossible in the private-tab model of AI. A shareable teammate requires an identity others can address, memory that persists beyond one person's history, work that lands where others can see it, and attribution, so everyone knows which edits were whose. Every one of those is a property of a shared workspace, not of a model. This is why sharing arrived years after chatbots did: the missing piece was never intelligence. It was [the roof](/blog/work-is-moving-to-where-your-agents-are).

There's one more requirement, easy to miss: someone has to stay accountable. A shared teammate still has one owner and one reporting line. Shared use, single accountability, the same arrangement as any internal service team, and the reason sharing doesn't dissolve into nobody-owns-it chaos.

## The compounding math

Here's what changes in the arithmetic of building a company. In the old model, capability arrived one hire at a time, and each hire helped whoever they reported to. In this model, one well-made teammate is a capability the entire company gains at once, and [every specialist anyone hires](/blog/hire-specialists-not-software) makes the shared pool deeper. Five people who each build one great teammate haven't added five tools. They've added five colleagues everyone has.

That's the quiet reason small teams running agents feel bigger than their headcount: their best hires don't queue, don't divide, and don't belong to a department. Hire once, staff everywhere.